ADC16_cross_corr_tests - david-macmahon/wiki_convert_test GitHub Wiki

Summary of cross talk measurements

  • Differential rev 1 (aka Samtec Vport rev 1)
    • These numbers apply when rms counts and digital gains are rms=28: 1, 2, 4, rms=14: 1, 2, 4 or rms=7: 1, 2, 4
  • Neighboring inputs on 1 chip with NO Samtec Vport cable:
    • -36 dB @ 50 MHz
  • Separated inputs on 1 chip:
    • no data
  • Inputs on 2 chips with Samtec Vport cable:
    • -39 dB @ 50 MHz
  • Inputs on 2 chips with NO Samtec Vport cable:
    • -50 dB @ 50 MHz (tested only at rms=28 and digital gain=2.)
  • Inputs on 2 boards:
    • no data
  • Differential rev 2
    • no data. This board was never fabricated let alone assembled.
  • RJ45 rev 1
    • These numbers apply when
      • input: wide band noise DC to 67 MHz (SLP70 LPF)
      • 3 ft CAT7 cable. (CAT6 and CAT5E are measurably worse)
      • rms counts is about 10 and ADC IC digital gain is 1, 2, 4 or 8.
    • inputs 1 v 3 on 1 ADC IC:
      • -32 dB
      • if ground inputs 2 and 4 at both ends of the CAT7 cable:-34 dB
    • inputs 1 v 4 on 1 ADC IC:
      • A1 v A4 : -31 dB
      • B1 v B4 : -32 dB
      • C1 v C4 : -32 dB
      • C1 v C4 and C2 and C3 grounded at both ends of the CAT7 cable:-33 dB
    • inputs 2 v 4 on 1 ADC IC
      • B2 v B4 : -39 dB
      • B2 v B4 and B1 and B3 grounded at both ends of the CAT7 cable: -46 dB (at 10MHz) to -40 dB (at 65 MHz)
    • Inputs on 2 chips with RJ45 adapter & CAT7s:
    • -49 dB @ 50 MHz
  • Inputs on 2 boards with RJ45 adapter & CAT7s:
    • no data
  • Coax rev 1
    • These numbers apply when rms counts and digital gains are rms=14: 1, 2, 4 and sometimes 8
    • neighboring ADC IC inputs 1
      • -34 dB @ 50 MHz
      • -28 and -34 dB @ 150 MHz (strong function of exact input pins)
    • Separated inputs on 1 chip:
      • -30 and -40 dB @ 150 MHz (strong function of exact input pins)
    • Separated inputs on 2 chips:
      • better than -41 dB @ 150 MHz
    • Separated inputs on 2 boards:
      • no data
  • Coax rev 2
    • No data

ADC16x250-8 differential rev 1

Lab setup A description of the test setup for the first 4 tests below may be found at test setup (PDF)

Please see noise_sources for a description of the RAL noise box used for these tests.

' Test 1: SLP70 into reworked balun into A1 versus SLP90 into reworked balun into A2'

summary plot file (PDF)

  • Digital Vgains of 1, 2, 4 and 8 look about the same. Gains above 8 produce quite different responses.
  • The Hittite ADC is pretty impressive; 3 of the 5 extra internal least significant bits seem usable at first inspection. Using the last 2 least significant bits will come with possibly significant tradeoffs.
  • These operating conditions seem about the same :
    • rms=28: dBms, digital Vgain (-1, 1), (-7, 2), (-13, 4) and (-19, 8)
    • rms=14: dBms, digital Vgain (-7, 1), (-13, 2) and (-19, 4)
    • rms= 7: dBms, digital Vgain (-13, 1) and (-19, 2)
    • rms= 3: dBms, digital Vgain (-19, 1)
  • The SLP90 filter is too wide bandwidth for 200 MHz sample rate. There is significant aliasing. This filter was used only because it was the closest available in house when the tests were performed.
  • test spreadsheet (PDF)
  • plot files (tar)
  • lab test notes (TXT)

' Test 2: SLP70 into reworked balun into A1 versus SLP70 into reworked balun into A2'

summary plot file (PDF)

  • Digital Vgains of 1, 2, 4 and 8 look about the same. Gains above 8 produce quite different responses.
  • The Hittite ADC is pretty impressive; 3 of the 5 extra internal least significant bits seem usable at first inspection. Using the last 2 least significant bits will come with possibly significant tradeoffs.
  • These operating conditions seem about the same :
    • rms=28: dBms, digital Vgain (-1, 1), (-7, 2) and (-13, 4). (-19,8) has some narrow spikes of nonsense.
    • rms=14: dBms, digital Vgain (-7, 1), (-13, 2) and (-19, 4)
    • rms= 7: dBms, digital Vgain (-13, 1) and (-19, 2). (-25,4) wasn't that much worse.
    • rms= 3: dBms, digital Vgain (-19, 1)
  • test spreadsheet (PDF)
  • plot files (tar)
  • lab test notes (TXT)

' Test 3: SLP70 into A3 versus SLP70 into D2'

summary plot file (PDF)

  • The tests with input power level of -19 dBm need to be redone. There was a lab-wide VAC power glitch that occurred in the middle of this test that wasn't detected until the entire set of tests were completed (and boards shipped off).
  • Digital Vgains of 1, 2, 4, 8 and 16 look about the same. Gains above 16 produce quite different responses.
  • The Hittite ADC is pretty impressive; 3 and maybe 4 of the 5 extra internal least significant bits seem usable at first inspection. Using the least significant bit will come with possibly significant tradeoffs.
  • These operating conditions seem about the same :
    • rms=28: dBms, digital Vgain (-1, 1), (-7, 2) and (-13, 4) and (-25,16) only has some narrow spikes of nonsense.
    • rms=14: dBms, digital Vgain (-7, 1), (-13, 2) and (-25, 8) and (-31, 16) only have some narrow spikes of nonsense.
    • rms= 7: dBms, digital Vgain (-13, 1),(-25, 4) and (-31,8) only has some narrow spikes of nonsense.
    • rms= 3: dBms, digital Vgain (-25, 2) and (-31, 4)
  • test spreadsheet (PDF)
  • plot files (tar)
  • lab test notes (TXT)

' Test 4: SMA Chunk filter into reworked balun into A2 versus SMA Chunk filter into D1'

deep integration with -7dBm digital Vgain 2 and rms 28 (PDF)

  • While a VLBI observations were running a large number of tests were run at 1 fixed setting to see how far down we could measure the crosstalk.
  • These inputs should be well isolated from each other as they enter different ICs via traces on different sides of the board over two different source cables.
  • The response looks pretty darn good at just better than 50 dB down.
  • SMA Chunk bandpass filters were used in place of the SLP70 low pass filters.
  • The SMA Chunk filter is being used as a poor simulation of a bandpass filter for the 2nd Nyquist band of a 200 MHz clock.
    • F1dB measured range is 107 to 201 MHz
    • F3dB measured range is 104 to 203.5 MHz
    • F6dB measured range is 102.2 to 205.2 MHz
    • F10dB measured range is 101 to 207.5 MHz
    • F20dB measured range is 98 to 215 MHz
    • F30dB measured range is 86.3 to 237 MHz

ADC16x250-8 differential rev 2

This board has never been fabricated let alone assembled. There are no test results to report.

ADC16x250-8 RJ45 rev 1

Please see noise_sources for a description of the RAL noise box used for these tests.

Please see SMA_to_RJ45_adapter for a description of the SMA to RJ45 adapter board used for these tests.

ADC16x250-8 coax rev 1

Lab setup

Please see noise_sources for a description of the RAL noise box used for some of these tests.

ADC16x250-8 coax rev 2

No data.

Test time ran out; the hardware had to be deployed in the Karoo before tests could be run. As this board is 1 somewhat subtle edit cycle of the coax rev 1 board it is assumed the measurements will be about the same for the rev 2 board. Actually the numbers should be slightly better for the rev 2 board because of greater physical separation between the baluns and other analog components and other PCB place and route changes.